Google’s TurboQuant has the internet joking about Pied Piper from HBO's "Silicon Valley." The compression algorithm promises to shrink AI’s “working memory” by up to 6x, but it’s still just a lab experiment for now.
Abstract
A probabilistic model is introduced for the problem of stimulating a
large male audience. Double jerking is considered, in which two shafts
may be stimulated with a single hand. Both tip-to-tip and shaft-to-shaft
configurations of audience members are analyzed. We demonstrate that
pre-sorting members of the audience according to both shaft girth and leg
length allows for more efficient stimulation. Simulations establish steady
rates of stimulation even as the variance of certain parameters is allowed
to grow, whereas naive unsorted schemes have increasingly flaccid perfor-
mance.
Someone already done a paper on it:
https://ia800308.us.archive.org/32/items/pdfy-tG1MuMpwvrML6QD0/228831637-Optimal-Tip-to-Tip-Efficiency.pdf
Abstract A probabilistic model is introduced for the problem of stimulating a large male audience. Double jerking is considered, in which two shafts may be stimulated with a single hand. Both tip-to-tip and shaft-to-shaft configurations of audience members are analyzed. We demonstrate that pre-sorting members of the audience according to both shaft girth and leg length allows for more efficient stimulation. Simulations establish steady rates of stimulation even as the variance of certain parameters is allowed to grow, whereas naive unsorted schemes have increasingly flaccid perfor- mance.